Carbon Emissions and Economic Growth in the Planting Industry: Evidence from China

Abstract

This study systematically analyzes the temporal variation characteristics, driving mechanisms, and decoupling relationship between carbon emissions and economic output in China’s planting industry. Using a dynamic panel model, LMDI decomposition, and coupling coordination model, it explores the main influencing factors of carbon emissions and their dynamic evolution. The findings reveal that from 2003 to 2022, carbon emissions in the planting industry exhibited a phased trend of rising first and then declining, with a limited overall reduction. Carbon emissions demonstrated significant path dependency. Planting industry output and agricultural investment were identified as the primary driving factors for carbon emissions, while energy intensity and mechanization levels had significant inhibitory effects. Decoupling analysis showed that weak decoupling dominates, with strong decoupling achieved only in specific regions and periods, highlighting significant regional disparities. Coupling coordination analysis indicated that the coordination between positive driving factors and carbon emissions improved annually, whereas the coordination related to rural electricity consumption declined in recent years. This study suggests that promoting precision agriculture and clean energy technologies, optimizing agricultural investment structures, implementing region-specific policies, and enhancing land resource planning can help us achieve the coordinated goals of high-quality agricultural development and carbon reduction. The findings provide theoretical insights and policy recommendations for low-carbon agricultural development and serve as a reference for global agricultural green transformation.

1. Introduction

Global climate warming has become a shared challenge faced by humanity, with excessive carbon emissions identified as its direct cause [1,2,3]. As the world’s largest developing country, China has made remarkable strides in agricultural production over the past few decades. However, this progress has not come without significant resource and environmental challenges. The increased demand for food has been met with extensive use of agricultural inputs such as pesticides, fertilizers, agricultural films, and machinery, which, while boosting production, have led to the over-exploitation of agricultural resources, degradation of ecosystems, and rising greenhouse gas emissions [4,5]. Agricultural carbon emissions pose serious environmental threats and are closely related to food security and sustainable agricultural development. Although agriculture is not the primary source of carbon emissions in China, its considerable scale makes its emissions significant, with substantial potential for reduction [6,7]. In this context, promoting greenhouse gas mitigation in agriculture is not just crucial for achieving China’s carbon neutrality goals, but is also an urgent need for optimizing agricultural structures, enhancing production efficiency, and advancing long-term sustainability in agricultural development. This imperative requires a robust theoretical framework integrating environmental economics principles—particularly, the Environmental Kuznets Curve (EKC) hypothesis, Decoupling Theory, and the Mechanization Paradox—to guide evidence-based policy formulation.

Current research on agricultural carbon emissions predominantly focuses on carbon accounting methods, including life cycle analysis (LCA) [8], input–output analysis [9], and the IPCC Inventory Method [10], which offer valuable insights into the sources and intensity of emissions in agricultural systems. However, these methods often fail to capture the dynamic evolution and regional heterogeneity of agricultural emissions. For example, LCA provides a comprehensive view of emissions across the entire life cycle of agricultural production but is data-intensive and overlooks underlying theoretical mechanisms governing agricultural emissions [11]. Existing studies, such as Liu et al., have emphasized the importance of sector-wide assessments but overlook temporal variations in emission drivers [12]. Similarly, national-level analyses have revealed aggregate trends in agricultural emissions but lack a detailed exploration of spatial disparities across different regions of China [13]. While some studies have explored emission sources, such as land use, consumption of agricultural inputs, livestock production, and the burning of agricultural residues, few have systematically examined emissions through the lens of established economic and environmental theories, such as the Environmental Kuznets Curve (EKC), Decoupling Theory, and Sustainable Innovation Frameworks [14]. To address this gap, this study establishes a theoretical foundation for understanding agricultural carbon emissions by incorporating key economic and sustainability theories. The Environmental Kuznets Curve (EKC) hypothesis suggests that carbon emissions initially rise with economic development but eventually decline as technological advancements and structural shifts reduce environmental impact [15]. Applying this framework to China’s agricultural sector, this study examines whether increased mechanization and efficiency improvements have facilitated an EKC-like turning point in carbon emissions. Additionally, this research engages with the Decoupling Theory to assess whether the planting industry in China has achieved relative or absolute decoupling of economic growth from emissions. While some studies have suggested a weak decoupling trend, the extent to which this is influenced by technological advancements, structural transformations, and policy interventions remains unclear.

Moreover, although national-level studies indicate a general decline in agricultural carbon emission intensity over the years, significant regional disparities remain [16]. These disparities are most evident when comparing emissions across China’s eastern, central, and western regions. The eastern region, with its advanced agricultural technologies and stronger policy enforcement, exhibits relatively lower carbon emissions, while the central and western regions, still dependent on more traditional and resource-intensive practices, experience higher emissions.

Despite significant progress, there is still a lack of comprehensive understanding of the multidimensional interactions and dynamic evolution of agricultural carbon emissions. Existing research often adopts a static perspective, overlooking key mechanisms such as the role of technological advancements and economic investments in shaping emissions trends [17]. While many studies have focused on the spatiotemporal characteristics and accounting of emissions, few have systematically tested hypotheses on the causal relationships between these factors. For example, does agricultural mechanization effectively reduce carbon emissions by improving production efficiency (H1)? Conversely, does increased agricultural investment lead to higher emissions due to resource-intensive expansion (H2)? Furthermore, the decoupling relationship between economic growth in the planting industry and carbon emissions has yet to be fully explored. Understanding these dynamics is critical for formulating effective policies for carbon reduction.

This study fills this gap by employing a dynamic panel model, Logarithmic Mean Divisia Index (LMDI) decomposition, and a coupling coordination model, enabling a comprehensive examination of the key drivers behind carbon emissions in the planting industry. The dynamic panel model allows for an in-depth exploration of emission path dependency, a key theoretical concept that suggests historical emission levels influence current and future trends due to long-term investment in agricultural infrastructure and mechanization.

Through this approach, this study aims to explore complex causal relationships between technological innovation, agricultural investment, mechanization, and carbon emissions, while also examining the regional and temporal patterns of the decoupling relationship between economic output in the planting industry and carbon emissions. By explicitly testing these relationships using panel econometric methods, this study provides empirical insights into the drivers of emission trends rather than merely describing their characteristics. Furthermore, by integrating these methodologies, the study highlights the evolution of carbon emissions across different regions and proposes tailored policy recommendations for optimizing carbon reduction strategies in China’s agricultural sector. Moreover, this study contributes to international research on sustainable agriculture by offering a robust methodological framework that can be applied to other developing countries facing similar challenges in reducing agricultural emissions.

One critical aspect of this research is its focus on regional disparities in carbon emissions. The eastern region’s success in mitigating emissions is largely attributed to factors such as technological advancements (e.g., precision agriculture, smart irrigation systems), strong policy frameworks (e.g., carbon trading programs, green subsidies), and superior infrastructure (e.g., high mechanization and electrification rates). However, the effectiveness of these policies has not been systematically evaluated. In contrast, the central and western regions, with less advanced agricultural technologies and fewer policy incentives, continue to face challenges in reducing emissions. For instance, while government subsidies have encouraged mechanization in the western region, it remains unclear whether these subsidies primarily promote low-carbon machinery or high-energy-consuming agricultural equipment. This study provides a detailed analysis of these regional differences, offering insights into the specific drivers behind each region’s emission trends.

In terms of methodology, this research employs a dynamic panel model to account for the temporal dependency and dynamic relationships between agricultural production and carbon emissions. The LMDI decomposition method quantifies the contributions of different factors to changes in carbon emissions, while the coupling coordination model assesses the synchronization between economic growth and environmental impact. These methodologies are particularly suitable for analyzing complex, multifactorial systems like agriculture, where emissions result from a combination of technological, economic, and policy factors. Furthermore, to enhance the robustness of our findings, additional analyses such as Granger causality tests and spatial econometric modeling (e.g., the Spatial Durbin Model) are conducted to explore both the causal structure and regional spillover effects of emission drivers.

2. Theoretical Framework

2.1. Decoupling Theory and the Dynamic Relationship of Agricultural Carbon Emissions

Decoupling Theory examines the relationship between economic growth and environmental impact, proposing that economic expansion can either continue to drive emissions (non-decoupling), result in relative decoupling (where emissions grow at a slower rate than economic output), or achieve absolute decoupling (where emissions decrease despite continued economic growth) [18,19,20]. In the agricultural sector, the decoupling of carbon emissions from economic growth is influenced by mechanization, energy efficiency, land use patterns, and policy interventions. Previous studies have shown that while some regions in China’s agricultural sector have exhibited relative decoupling, absolute decoupling remains largely unachieved [21,22]. This study employs decoupling analysis to assess the evolving relationship between economic expansion and carbon emissions in the planting industry, identifying regional disparities and key transition phases in the low-carbon development of agriculture.

2.2. Spatial Effects and Regional Distribution of Carbon Emissions

Carbon emissions in agriculture are not only determined by local economic activities but also influenced by spillover effects from adjacent regions. These spatial dependencies can lead to clustering or diffusion patterns in carbon emissions across geographic areas [23,24]. Empirical studies indicate significant spatial autocorrelation in agricultural carbon emissions, particularly in China’s economically developed regions and major agricultural production zones [25]. This study applies spatial autocorrelation analysis to detect spatial patterns in agricultural carbon emissions and further employs the Spatial Durbin Model (SDM) to quantify spillover effects, thereby capturing the inter-regional dynamics of agricultural carbon emissions.

2.3. Decomposing the Driving Forces of Carbon Emissions: The LMDI Model

The Logarithmic Mean Divisia Index (LMDI) model is a widely used method for decomposing changes in carbon emissions into their contributing factors [26]. It allows for a comprehensive examination of the role of energy intensity, economic structure, agricultural inputs, and technological improvements in driving emissions. Previous studies have applied the LMDI to quantify the impact of mechanization, fertilizer use, and energy consumption on agricultural carbon emissions [27]. In this study, the LMDI model is used to systematically attribute changes in carbon emissions to specific influencing factors, providing empirical evidence on the effectiveness of policy interventions and technological advancements in reducing emissions.

2.4. Coordinating Economic Growth and Carbon Emissions: The Coupling Coordination Model

Coupling coordination theory is widely used to evaluate the interactions between economic development and environmental sustainability [28]. This study applies it to examine the coordination between agricultural economic growth and carbon emissions across different regions. Research has demonstrated that there are significant regional variations in the coupling coordination of economic and environmental factors in China. Coastal provinces with advanced agricultural technologies and strong policy incentives tend to exhibit higher coordination, while central and western provinces remain in a state of imbalance due to their continued reliance on resource-intensive agricultural practices [29]. By constructing a coupling coordination model, this study assesses the evolving relationship between economic growth and carbon emissions in the planting industry and identifies regions where policy interventions are needed to enhance sustainable development.

2.5. Dynamic Evolution of Carbon Emissions: The Role of Path Dependence

Carbon emissions in agriculture exhibit path dependence, meaning that historical emissions significantly influence current and future trends. This long-term persistence is shaped by factors such as agricultural infrastructure investments, energy consumption patterns, and policy frameworks [30]. To analyze the dynamic evolution of carbon emissions, this study employs the dynamic panel model (DPM), which incorporates lagged variables to account for the persistence of emissions over time. Prior research has used dynamic panel regression to assess the impacts of energy consumption, mechanization, and government subsidies on carbon emissions [31]. By adopting this approach, this study enhances the robustness of its empirical analysis and provides a more precise assessment of the long-term determinants of carbon emissions in China’s planting industry.

2.6. Constructing the Analytical Framework

This study integrates Decoupling Theory, spatial autocorrelation, the LMDI model, the coupling coordination model, and the dynamic panel model to establish a comprehensive analytical framework for understanding the evolution of agricultural carbon emissions in China. The decoupling analysis identifies whether economic growth in the planting industry is accompanied by emission reductions, capturing the extent to which low-carbon transitions are taking place. Spatial autocorrelation analysis explores regional clustering effects and spatial spillover mechanisms, revealing how emissions in one region influence those in neighboring areas. The LMDI decomposition method disaggregates carbon emissions into their key drivers, quantifying the contributions of mechanization, energy efficiency, and structural transformations. The coupling coordination model measures the synchronization between agricultural development and carbon emissions, assessing whether different regions are on sustainable development pathways. Finally, the dynamic panel model accounts for the path-dependent nature of carbon emissions, controlling for historical influences and providing a robust econometric foundation for evaluating policy effectiveness. By integrating these theoretical perspectives, this study constructs a systematic framework for analyzing the drivers, regional disparities, and policy implications of carbon emissions in the planting industry.

3. Data Sources and Research Methods

3.1. Data Sources

The data used in this study, including fertilizer application (converted to pure nutrients), agricultural film consumption, pesticide consumption, diesel consumption by agricultural machinery, crop sown area, effective irrigation area, total agricultural output value, cultivated land area, and agricultural labor force, were sourced from the China Rural Statistical Yearbook (2004–2023) and statistical yearbooks covering 31 provincial-level administrative regions in China. Data on agricultural water use were obtained from the China Environmental Statistical Yearbook (2004–2023).

To address missing values, this study initially employed linear interpolation. However, considering its potential biases in capturing nonlinear trends, we further applied the multiple imputation by chained equations (MICE) method and compared the results with those obtained using random forest regression-based imputation. The robustness test indicated that MICE provided more stable estimates for variables with moderate missing rates (below 10%), making it the preferred method for data imputation in this study.

3.2. Measurement of Carbon Emissions in the Planting Industry

Carbon emissions in agriculture primarily originate from land use activities, the consumption of agricultural inputs, and energy utilization. It is generally accepted that the carbon emissions from farmland production activities mainly stem from direct and indirect carbon emissions caused by water resource exploitation, land development, and fuel consumption. These include the use of fertilizers, pesticides, agricultural film, diesel for agricultural machinery, irrigation, and tillage. Due to regional differences in crop types and planting areas, as well as challenges in data acquisition, this study does not consider carbon emissions arising from the growth and development processes of crops.

Following previous research, this study employs the carbon emission coefficient method provided by the Intergovernmental Panel on Climate Change (IPCC) to estimate agricultural carbon emissions in China’s 31 provinces. Despite the availability of alternative estimation approaches, such as life cycle assessment (LCA) and input–output analysis, these methods are often constrained by data accessibility and sector-specific assumptions, making the coefficient-based approach a more practical choice for large-scale regional studies. The carbon emission calculation formula is as follows:

$$E_T=E_f+E_p+E_m+E_e+E_i$$

(1)

where $E_T$ represents the total agricultural carbon emissions, $E_f$ denotes carbon emissions from fertilizer use, $E_p$ represents carbon emissions from pesticide use, $E_m$ is carbon emissions from agricultural plastic film use, $E_e$ refers to carbon emissions from agricultural machinery use, and $E_i$ represents carbon emissions from agricultural irrigation. Each source’s carbon emissions are calculated using the following formulas:

$$E_f={ G}_f\times a$$

(2)

$$E_p=G_p\times b$$

(3)

$$E_m=G_m\times c$$

(4)

$$E_e=G_e\times d+A_e\times g$$

(5)

$$E_i=A_i\times h$$

(6)

where $G_f$ is the fertilizer consumption (a = 0.8956 kg/kg), $G_p$ is the pesticide consumption (b = 4.9341 kg/kg), $G_m$ is the agricultural plastic film consumption (c = 5.18 kg/kg), $G_e$ is the total agricultural machinery power (d = 0.18 kg/kW), $A_e$ is the farmland area (g = 16.47 kg/hm²), and $A_i$ is the irrigated agricultural area (h = 266.48 kg/hm²). The coefficients a, b, c, d, g, h represent the carbon emission factors for each source, with values derived from the IPCC guidelines and relevant domestic and international studies. A sensitivity analysis was conducted to test the robustness of these coefficients against varying input levels, confirming the reliability of the selected values.

3.3. Trend of Changes in Carbon Emission Intensity in the Planting Industry

The Slope method is commonly used to evaluate the variation trend of variables over time [32]. This study applies the Slope method to assess the variation trend and rate of change in carbon emissions from the planting industry during 2003–2022, using the following formula:

$$\mathrm{S}\mathrm{l}\mathrm{o}\mathrm{p}\mathrm{e}={\displaystyle \frac{t\times \sum _{i=1}^t{x}_i{CI}_i-\sum _{i=1}^t{x}_i\sum _{i=1}^t{CI}_i}{t\times \sum _{i=1}^t{x}_i^2-{ \left(\sum _{i=1}^t{x}_i\right)}^2}}$$

(7)

$${CI}_i=C_i/E_i$$

(8)

In the formula, t represents the total number of years from 2003 to 2022; $x_i$ denotes the i-th year (with 2003 as the first year); $C_i$ represents the carbon emissions of the planting industry in the i-th year; ${CI}_i$ indicates the carbon emission intensity of the planting industry in the i-th year; and $E_i$ denotes the economic output of the planting industry in the i-th year. When the Slope > 0, it indicates that the carbon emissions of the planting industry have increased annually. Conversely, when the Slope < 0, it signifies that carbon emission intensity has decreased annually. This study employs the standard deviation classification method to categorize the trends in carbon emissions across regions into four types. The specific classification criteria are shown in Table 1.

Table 1. Criteria for classification of changing trend types ¹ [33].

Growth Type	Slow Growth	Moderate Growth	Rapid Growth	Surge Growth
Slope Value	$<\overline{\mathrm{x}}$ − 0.5 s	$\overline{\mathrm{x}}$$-0.5 \mathrm{s}~\overline{\mathrm{x}}$+ 0.5 s	$\overline{\mathrm{x}}$ + $0.5 \mathrm{s}~\overline{\mathrm{x}}$ + 1.5 s	$>\overline{\mathrm{x}}$ + 1.5 s

¹ $\overline{\mathrm{x}}$ and s represent the mean and standard deviation of the Slope values for all regions, respectively.

Table 1. Criteria for classification of changing trend types ¹ [33].

Growth Type	Slow Growth	Moderate Growth	Rapid Growth	Surge Growth
Slope Value	$<\overline{\mathrm{x}}$ − 0.5 s	$\overline{\mathrm{x}}$$-0.5 \mathrm{s}~\overline{\mathrm{x}}$+ 0.5 s	$\overline{\mathrm{x}}$ + $0.5 \mathrm{s}~\overline{\mathrm{x}}$ + 1.5 s	$>\overline{\mathrm{x}}$ + 1.5 s

¹ $\overline{\mathrm{x}}$ and s represent the mean and standard deviation of the Slope values for all regions, respectively.

The classification thresholds are based on the mean ($\overline{\mathrm{x}}$) and standard deviation (s) of the computed Slope values across all provinces. Regions with Slope values lower than $\overline{\mathrm{x}}$ − 0.5 s are classified as slow growth, indicating a below-average increase in carbon emission intensity, which may result from policy-driven mitigation efforts, technological innovation, or efficiency improvements in agricultural production. Moderate growth regions fall between $\overline{\mathrm{x}}$ − 0.5 s and $\overline{\mathrm{x}}$ + 0.5 s, reflecting a trend that aligns with the national average, where economic expansion and environmental regulation interact in a relatively balanced manner. When Slope values range from $\overline{\mathrm{x}}$ + 0.5 s to $\overline{\mathrm{x}}$ + 1.5 s, the region is categorized as rapid growth, signifying a significant rise in carbon emission intensity, which is often driven by increased agricultural inputs, expanded mechanization, and inefficient resource utilization. If the Slope exceeds $\overline{\mathrm{x}}$ + 1.5 s, the region falls under surge growth, representing an extreme increase in emission intensity that is likely associated with large-scale land expansion, excessive reliance on fertilizers, and energy-intensive farming methods. This classification framework provides a systematic approach to analyzing emission trends across different regions, helping identify areas that require urgent policy intervention for emission reduction.

3.4. Decoupling Relationship Between Agricultural Carbon Emissions and Economic Development

To analyze the decoupling relationship between carbon emissions in the planting industry and its economic output, this study adopts the Tapio decoupling model [34,35]. This model effectively evaluates the relationship between carbon emissions and economic growth and has strong explanatory power. The basic formula of the Tapio decoupling model is as follows:

$$D={\displaystyle \frac{∆C}{∆E}}$$

(9)

where D represents the decoupling elasticity coefficient, ΔC denotes the growth rate of carbon emissions in the planting industry, and ΔE indicates the growth rate of economic output in the planting industry. The growth rates are calculated using the following formulas:

$$∆C={\displaystyle \frac{C_t-{ C}_{2003}}{C_{2003}}}$$

(10)

$$∆E ={\displaystyle \frac{E_t-{ E}_{2003}}{E_{2003}}}$$

(11)

where $C_{2003}$ and $C_t$ represent the carbon emissions of the planting industry in 2003 and year t, respectively, while $E_{2003}$ and $E_t$ denote the economic output of the planting industry in 2003 and year t, respectively. When determining the decoupling type between carbon emissions and economic output in the planting industry, it is necessary to comprehensively consider the positive or negative relationships of the growth rates for carbon emissions and economic output, along with the decoupling elasticity coefficient. The specific classification criteria for decoupling types are shown in Table 2.

Table 2. Classification criteria for decoupling types [36,37].

$∆\mathit{C}$	$∆\mathit{E}$	D	Decoupling Type	Code
$∆C$ < 0	$∆E$ > 0	D < 0	Strong Decoupling	1
$∆C$ > 0	$∆E$ > 0	0 < D < 0.8	Weak Decoupling	2
$∆C$ > 0	$∆E$ > 0	0.8 ≤ D ≤ 1.2	Expansion Coupling	3
$∆C$ > 0	$∆E$ > 0	0 ≤ D ≤ 1.2	Weak Negative Decoupling	4
$∆C$ > 0	$∆E$ > 0	D > 1.2	Negative Decoupling	5
$∆C$ < 0	$∆E$ < 0	D > 0	Recession Coupling	6
$∆C$ > 0	$∆E$ < 0	D < 0	Recession Negative Decoupling	7

Table 2. Classification criteria for decoupling types [36,37].

$∆\mathit{C}$	$∆\mathit{E}$	D	Decoupling Type	Code
$∆C$ < 0	$∆E$ > 0	D < 0	Strong Decoupling	1
$∆C$ > 0	$∆E$ > 0	0 < D < 0.8	Weak Decoupling	2
$∆C$ > 0	$∆E$ > 0	0.8 ≤ D ≤ 1.2	Expansion Coupling	3
$∆C$ > 0	$∆E$ > 0	0 ≤ D ≤ 1.2	Weak Negative Decoupling	4
$∆C$ > 0	$∆E$ > 0	D > 1.2	Negative Decoupling	5
$∆C$ < 0	$∆E$ < 0	D > 0	Recession Coupling	6
$∆C$ > 0	$∆E$ < 0	D < 0	Recession Negative Decoupling	7

3.5. Dynamic Panel Model

To explore the primary driving factors influencing carbon emissions in the planting industry, this study selects seven key explanatory variables: energy consumption, the gross output value of agriculture, forestry, animal husbandry, and fishery (AFAF output), planting industry output, rural population, total agricultural machinery power, cultivated land area, rural electricity consumption, and agricultural investment. Energy consumption is a direct source of carbon emissions and reflects energy efficiency and emission levels during agricultural production. AFAF output and planting industry output represent overall agricultural economic development and the scale of the planting industry, respectively, both of which significantly impact changes in carbon emissions. The rural population, as a key component of the labor force, influences labor input and the transformation of production methods in agriculture. Total agricultural machinery power is an important indicator of agricultural mechanization levels, closely related to production efficiency and energy consumption. The cultivated land area serves as the spatial foundation for agricultural production, directly affecting the production scale and carbon emission intensity of the planting industry. Rural electricity consumption reflects the level of agricultural electrification and changes in the energy structure. Finally, agricultural investment plays a critical role in advancing agricultural technological progress and optimizing production structures, exerting a long-term regulatory effect on carbon emissions.

This study employs the Generalized Method of Moments (GMM) to construct a dynamic panel model [38], aiming to reveal the dynamic mechanisms through which these factors influence carbon emissions in the planting industry. The basic form of the model is as follows:

$$\ln C_{it}=\alpha \ln C_{i,t-1}+{\beta }_1\ln E_{it}+{\beta }_2\ln A_{it}+{\beta }_3\ln P_{it}+{\beta }_4\ln M_{it}+{\beta }_5\ln L_{it}+{\beta }_6\ln U_{it}+{\beta }_7\ln I_{it}+{\gamma }_i+{\epsilon }_{it}$$

(12)

where $\ln C_{it}$ represents the carbon emissions of the planting industry in province i and year t, while $\ln C_{i,t-1}$ is the lagged term capturing path dependency and dynamic effects. The explanatory variables include $\ln E_{it}$ (energy consumption), $\ln A_{it}$ (AFAF output), $\ln P_{it}$ (planting industry output), $\ln M_{it}$ (total agricultural machinery power), $\ln L_{it}$ (cultivated land area), $\ln U_{it}$ (rural electricity consumption), and l$\ln I_{it}$ (agricultural investment). ${\gamma }_i$ denotes province-specific fixed effects to control for unobservable heterogeneity, and ${\epsilon }_{it}$ is the random error term. The parameters α and β1, β2, …, β7 are coefficients to be estimated.

To eliminate the influence of province-specific fixed effects, the model is first transformed into its first-difference form. This step controls for unobservable heterogeneity across provinces. The first-differenced model is expressed as follows:

$$\ln C_{it}=\alpha ∆\ln C_{i,t-1}+{\beta }_1∆\ln E_{it}+{ \beta }_2∆\ln A_{it}+{\beta }_3∆\ln P_{it}+{\beta }_4∆\ln M_{it}+{\beta }_5∆\ln L_{it}+{\beta }_6∆\ln U_{it}+{\beta }_7∆\ln I_{it}+∆{\epsilon }_{it}$$

(13)

Lagged values of the variables are used as instrumental variables, along with the introduction of exogenous instrumental variables to address the endogeneity issue in the model. Based on the moment conditions constructed with these instruments, the system GMM method is employed to estimate the model parameters. The validity of the instrumental variables is evaluated using the Hansen test. Additionally, the first-order and second-order autocorrelation tests are conducted to ensure the appropriateness of the model specification. Ultimately, the estimation results quantify the elasticity of each variable’s impact on carbon emissions in the planting industry, providing empirical support for formulating policies to optimize agricultural carbon reduction strategies. In addition to GMM estimation, Granger causality tests are commonly used to examine the directional relationship between economic and environmental variables. This method determines whether past values of one variable can predict another variable’s future values. However, given that carbon emissions and economic factors are influenced by multiple dynamic forces, and endogeneity concerns are already addressed in the GMM framework, Granger causality tests serve as a supplementary check rather than a primary estimation tool.

As a robustness check, panel Granger causality tests were conducted on selected explanatory variables (e.g., agricultural mechanization, investment, and rural electricity consumption). The results confirm that mechanization significantly causes carbon emission reductions (p < 0.05), while agricultural investment exhibits bidirectional causality with emissions (p < 0.1), highlighting the complex interplay between technological progress and carbon dynamics.

3.6. Spatial Econometric Analysis

Carbon emissions in agriculture are not confined to individual provinces but often exhibit spatial dependence due to regional policy diffusion, resource distribution, and industrial linkages. Thus, traditional panel models may not fully capture the spillover effects of carbon emissions across regions. To address this issue, this study employs a Spatial Durbin Model (SDM), which extends spatial econometrics by incorporating both spatial lag effects and spillover effects of explanatory variables. The model is expressed as follows:

$$C_{it}=\rho W{C}_{it}+\sum _{k=1}^n{\beta }_k{X}_{k,it}+\sum _{k=1}^n{\theta }_k{WX}_{k,it}+{\mu }_i+{ϵ}_{it}$$

(14)

In the formula, $C_{it}$ represents the carbon emissions of province i at time t. $W{C}_{it}$ is the spatially lagged dependent variable, capturing the influence of neighboring regions’ carbon emissions. ${WX}_{k,it}$ represents the spatial spillover effects of key explanatory variables, such as mechanization and investment. $\rho$ is the spatial autoregressive coefficient, indicating the degree of spatial dependence. ${\beta }_k$ and ${\theta }_k$ are the coefficients for local and spillover effects, respectively. ${\mu }_i$ denotes province-specific fixed effects, and ${ϵ}_{it}$ is the error term. To verify the necessity of spatial modeling, we conducted Moran’s I test to examine the spatial autocorrelation of agricultural carbon emissions. The results confirm significant positive spatial correlation (p < 0.01), justifying the use of SDM.

3.7. LMDI Model

To further investigate the driving factors behind changes in carbon emissions in the planting industry, this study employs the Logarithmic Mean Divisia Index (LMDI) decomposition method. The LMDI has advantages such as additivity and adaptability to zero values, making it an effective tool for decomposing carbon emission changes and quantifying the contributions of various driving factors [12]. Based on the key variables in this study, the carbon emissions in the planting industry are decomposed into the driving effects of energy consumption, energy intensity, gross output value of agriculture, forestry, animal husbandry, and fishery (AFAF output), planting industry output, rural population, total agricultural machinery power, cultivated land area, rural electricity consumption, and agricultural investment. The decomposition model is as follows:

$$∆\mathrm{C}=∆C_E+∆C_A+∆C_P+∆C_M+∆C_L+∆C_U+∆C_I+∆C_H+∆C_K$$

(15)

$$∆C_X=L\left(C_t,C_{t-1}\right)·\ln \left({\displaystyle \frac{X_t}{X_{t-1}}}\right), X\in \left\{E,A,P,M,L,U,I,H,K\right\}$$

(16)

$$L\left(C_t,C_{t-1}\right)={\displaystyle \frac{C_t-C_{t-1}}{\ln \left(C_t\right)-\ln \left(C_{t-1}\right)}}$$

(17)

In the formula, $L\left(C_t,C_{t-1}\right)$ represents the logarithmic mean weight, ensuring additivity and eliminating the zero-value problem during decomposition. The components of carbon emission changes are as follows: $∆C_E$ denotes the energy consumption effect, reflecting the impact of changes in energy consumption on carbon emissions; $∆C_H$ represents the energy intensity effect, indicating the influence of changes in energy efficiency on carbon emissions; $∆C_A$ is the AFAF output effect, representing the impact of changes in the overall agricultural economic scale on carbon emissions; $∆C_P$ is the planting industry output effect, reflecting the driving influence of economic growth in the planting industry on carbon emissions; $∆C_K$ denotes the population size effect, capturing the impact of changes in the rural population on carbon emissions; $∆C_M$ represents the agricultural mechanization effect, indicating the influence of changes in total agricultural machinery power on carbon emissions; $∆C_L$ is the cultivated land area effect, reflecting the impact of changes in planting area on carbon emissions; $∆C_U$ denotes the rural electricity consumption effect, representing the influence of energy structure and electricity consumption changes on carbon emissions; and $∆C_I$ represents the agricultural investment effect, measuring the role of changes in agricultural infrastructure construction and technological input in carbon emissions.

3.8. Coupling Coordination Model

After identifying the key factors that significantly promote carbon emissions in the planting industry using the LMDI model, this study further employs a coupling coordination model to quantitatively analyze the coupling coordination relationships between these factors and carbon emissions. The analysis encompasses three dimensions: the coupling coordination relationships between each individual factor and carbon emissions, the overall coupling coordination relationships between all factors and carbon emissions, and the coupling coordination relationships among the factors themselves. The coupling coordination model is based on system coupling theory and incorporates a coordination degree model to quantitatively evaluate the strength of interactions and the level of coordination between two or more systems [39,40]. The coupling coordination relationships between carbon emissions and the driving factors in the planting industry are expressed as follows:

$$Z_{ij}={\displaystyle \frac{\sqrt[2]{U_i·U_j}}{U_i+U_j}}$$

(18)

$$T_{ij}=\alpha U_i+\beta U_j$$

(19)

$$Q_{ij}=\sqrt{Z_{ij}·T_{ij}}$$

(20)

Here, $Z_{ij}$ represents the coupling degree between carbon emissions and a driving factor; $U_i$ and $U_j$ are the comprehensive evaluation values of the two systems, which are calculated as the weighted average of normalized scores for their respective indicators. $Q_{ij}$ is the coordination degree between carbon emissions and the driving factor, while $T_{ij}$ is the weighted sum of the two systems’ comprehensive evaluation values. The weights α and β are typically set to 0.5 to reflect the balanced contribution of both systems. To quantitatively evaluate the coordination development status between systems, this study adopts a coupling coordination degree indicator to classify the level of coordination. Based on the value range of the coupling coordination degree, it is divided into 10 levels, each corresponding to a specific coordination type and level, as shown in

Table 3. Classification criteria for coupling coordination degree.

Coupling Coordination Degree Range (D)	Coordination Level	Coordination Degree
[0.0~0.1)	1	Extreme Imbalance
[0.1~0.2)	2	Severe Imbalance
[0.2~0.3)	3	Moderate Imbalance
[0.3~0.4)	4	Mild Imbalance
[0.4~0.5)	5	Near Imbalance
[0.5~0.6)	6	Barely Coordinated
[0.6~0.7)	7	Primary Coordination
[0.7~0.8)	8	Intermediate Coordination
[0.8~0.9)	9	Good Coordination
[0.9~1.0]	10	Optimal Coordination

.

3.9. Regional Division

Based on China’s official classification standards, the country is divided into three regions for this study: the eastern, central, and western regions. The eastern region includes Beijing, Tianjin, Hebei, Liaoning, Shanghai, Zhejiang, Jiangsu, Fujian, Shandong, Guangdong, and Hainan. The central region comprises Shanxi, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, and Hunan. The western region consists of Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Tibet, Shaanxi, Gansu, Qinghai, Ningxia, and Xinjiang (Figure 1).

4. Results

4.1. Carbon Emissions in the Planting Industry

At the national level, carbon emissions in the planting industry from 2003 to 2022 exhibited a trend of initial increase followed by a decrease. The total emissions rose steadily from 78.21 million tons in 2003 to a peak of 106.92 million tons in 2015, and then gradually declined to 89.99 million tons in 2022. This trend reflects significant fluctuations in planting activities during this period. The annual average carbon emissions reached 96.00 million tons. The increase from 2003 to 2015 was primarily driven by the demand for higher grain yields, enhanced mechanization, and expansion of planting areas. In contrast, the reduction in emissions after 2015 can be attributed to the continuous implementation of national energy conservation and emission reduction policies, advancements in agricultural technologies, and the promotion of resource-efficient agricultural practices. Additionally, the COVID-19 pandemic in 2020 temporarily disrupted agricultural production and supply chains, which may have led to short-term fluctuations in emissions. The restrictions on labor movement and transportation likely contributed to a temporary decline in certain agricultural activities, while the increased adoption of automated machinery and digital farming to compensate for labor shortages may have had long-term implications for emission trends. Despite the post-peak mitigation, the emission level in 2022 remained significantly higher than in the initial period, indicating that a full decoupling between agricultural production and carbon emissions has yet to be achieved (Figure 2).

At the regional level, carbon emissions in the eastern, central, and western regions exhibited significant spatial disparities and dynamic characteristics. The eastern region had an annual carbon emission average of 37.77 million tons, showing an overall declining trend. Emissions decreased from 34.69 million tons in 2003 to 31.88 million tons in 2022, a total reduction of 4.74 million tons. This trend reflects the higher degree of agricultural modernization in the eastern region, where technological upgrades and the adoption of resource-efficient agricultural practices effectively reduced carbon emissions. Particularly in economically advanced provinces such as Jiangsu, Zhejiang, and Shandong, adjustments in planting structure and progress in agricultural technologies have significantly reduced carbon intensity.

In comparison, the central region had an annual carbon emission average of 38.56 million tons, with total emissions increasing from 28.97 million tons in 2003 to 37.02 million tons in 2022, a growth of 6.16 million tons. This consistent increase is likely driven by the gradual improvement in agricultural infrastructure, intensified grain production tasks, and increased mechanization and fertilizer usage in the central region. However, the impact of the COVID-19 pandemic was also evident in this region, as temporary disruptions in agricultural supply chains may have slowed the rate of emission growth in 2020 and 2021. Although mechanization played a crucial role in mitigating labor shortages, fluctuations in input availability and market demand could have influenced planting decisions and production efficiency.

The western region reported an annual carbon emission average of 20.60 million tons, with total emissions rising from 14.55 million tons in 2003 to 21.10 million tons in 2022, an increase of 6.03 million tons. The growth in carbon emissions in the western region is primarily attributed to the expansion of agricultural production, increased irrigation demands, and intensified land reclamation activities for planting, especially in areas such as Xinjiang and Shaanxi, where these trends are particularly prominent. The pandemic-related labor shortages in some western provinces may have led to greater reliance on mechanized farming and remote monitoring technologies, which could influence future emission trends by altering the energy consumption structure in agricultural production.

At the provincial level, Shandong Province, the Inner Mongolia Autonomous Region, and the Xinjiang Uyghur Autonomous Region exhibited distinct heterogeneity. Shandong, as a major agricultural province, reported an annual carbon emission average of 8.62 million tons, with total emissions decreasing from 8.60 million tons in 2003 to 6.98 million tons in 2022, a reduction of 1.78 million tons. This reduction can be attributed to ongoing efforts in agricultural technology application, energy-saving policy implementation, and optimization of the planting structure in the province. Inner Mongolia reported an annual carbon emission average of 3.41 million tons, with total emissions increasing from 1.96 million tons in 2003 to 4.12 million tons in 2022, an increase of 1.98 million tons. The growth in Inner Mongolia’s emissions is closely related to the expansion of its farming and livestock production, intensified land reclamation activities, and the adoption of high-intensity agricultural practices. Xinjiang had the highest increase among the three regions, with an annual carbon emission average of 4.52 million tons. Total emissions rose from 2.47 million tons in 2003 to 5.73 million tons in 2022, an increase of 3.13 million tons. The expansion of agricultural production and high dependence on irrigation resources were the main drivers of carbon emission growth in Xinjiang. The COVID-19 pandemic also had an indirect effect on agricultural production in Inner Mongolia and Xinjiang, where disruptions in agricultural labor availability and input supplies influenced production strategies. Some areas experienced temporary reductions in emissions due to decreased agricultural activity, while others saw an acceleration in mechanization as a long-term response to labor shortages.

In summary, carbon emissions in the planting industry exhibited significant temporal and spatial heterogeneity across the country. The eastern region achieved notable emission reductions, while the central and western regions experienced continuous growth in emissions, reflecting differences in agricultural development stages, technological levels, and policy implementation effectiveness. Furthermore, the COVID-19 pandemic introduced additional complexities into these trends, causing short-term fluctuations in agricultural emissions due to disruptions in labor availability, input supply, and market demand. However, the pandemic also acted as a catalyst for increased mechanization and digitalization in agriculture, which may have long-term implications for emission reduction strategies. This regional and provincial spatial heterogeneity provides critical clues for studying the driving mechanisms of carbon emissions in the planting industry and underscores the importance of precise policy interventions to promote coordinated emission reduction across the country.

4.2. Temporal Evolution of Carbon Emission Intensity in the Planting Industry

From the perspective of the growth rate of carbon emission intensity in the planting industry (represented by the Slope value), significant spatial disparities were observed across regions. Based on the magnitude and rate of change of the Slope value, the trends can be categorized into four types: “slow growth”, “moderate growth”, “rapid growth”, and “surge growth”.

At the national level, the average Slope value was 0.3256. Regions classified as “slow growth” had an average Slope value of 0.2171, including Beijing, Liaoning Province, and Shanghai. These areas exhibited relatively slow growth in carbon emission intensity, likely due to their smaller planting industry scale, higher agricultural technology levels, and effective implementation of energy-saving and emission reduction policies. “Moderate growth” was the most common trend type, with an average Slope value of 0.3114, observed in typical regions such as Zhejiang Province, Fujian Province, and Hunan Province. These regions maintained stable growth in agricultural activities, with moderate increases in carbon emission intensity (

Table 4. Slope values of carbon emission intensity in the planting industry in China (2003–2022).

Region	Slope	Region	Slope
Eastern region	0.3215	Anhui	0.3875
Beijing	0.1894	Jiangxi	0.3502
Tianjin	0.3609	Henan	0.3156
Hebei	0.4302	Hubei	0.3561
Liaoning	0.2750	Hunan	0.2823
Shanghai	0.2183	Western region	0.3240
Jiangsu	0.3258	Chongqing	0.2874
Zhejiang	0.3134	Sichuan	0.2898
Fujian	0.3128	Guizhou	0.3247
Shandong	0.3183	Yunan	0.3559
Guangdong	0.2234	Tibet	0.1773
Guangxi	0.3311	Shaanxi	0.3785
Hainan	0.2741	Gansu	0.2973
Central region	0.3272	Qinghai	0.3567
Shanxi	0.3731	Ningxia	0.4929
Inner Mongolia	0.2876	Xinjiang	0.3090
Jilin	0.1625	China	0.3256
Heilongjiang	0.3595

).

Regions classified as “rapid growth” had an average Slope value of 0.3643, representing areas such as Inner Mongolia Autonomous Region, Heilongjiang Province, and Guizhou Province, where agricultural production expansion is more pronounced. The accelerated growth in carbon emission intensity in these regions may have resulted from the advancement of agricultural mechanization and further exploitation of land resources. “Surge growth” was concentrated in Hebei Province and Xinjiang Uyghur Autonomous Region, where the Slope value reached 0.4615. These regions experienced extremely rapid increases in carbon emission intensity, likely driven by significant expansion in planting areas, increased utilization of irrigation resources, and high-intensity agricultural production activities (Figure 3).

Overall, the growth rate of carbon emission intensity in the planting industry in China shows clear regional disparities. The classification of these trend types not only reveals differences in agricultural development levels and technological conditions across regions but also provides a reference for formulating region-specific agricultural carbon reduction policies.

4.3. Decoupling Types

Using 2003 as the base year, the decoupling relationship between carbon emissions and output value in the planting industry nationwide from 2004 to 2022 was predominantly characterized by “weak decoupling”, accounting for 83.3% (583 instances) of the total sample. This indicates that in most regions, the growth rate of carbon emissions was lower than that of output value, but significant decoupling had not yet been achieved. The “strong decoupling” type accounted for 9.8% (69 instances), primarily occurring in regions and periods with rapid advancements in agricultural technology or significant policy interventions, demonstrating a more evident decoupling between carbon emissions and output value. In contrast, types such as “expansion coupling”, “strong negative decoupling”, and “weak negative decoupling” were relatively rare, appearing 8, 3, and 2 times, respectively. However, these types highlight specific conditions in certain regions and years.

The “expansion coupling” type primarily reflects a state of synchronous growth in carbon emissions and output value, mostly concentrated in the early stages of rapid growth in the planting industry. Typical regions include Tianjin (2005), Hebei (2004 and 2005), Hainan (2004, 2005, and 2007), Tibet (2004 and 2005), and Jilin (2004). In these regions, the rapid expansion of planting activities in the early stages led to simultaneous increases in carbon emissions and output value. Additionally, the “strong negative decoupling” and “weak negative decoupling” types represent cases of dual decline in carbon emissions and output value, primarily occurring in Tianjin and Hainan, usually associated with economic fluctuations or adjustments in agricultural production activities.

From a temporal perspective, “weak decoupling” was the dominant type from 2004 to 2015, indicating that during this period, the growth rate of carbon emissions in most regions gradually fell below the growth rate of output value, but complete decoupling had not been achieved. From 2016, the frequency of the “strong decoupling” type increased, particularly prominent during 2020–2022, reflecting the significant impact of energy-saving policies and the promotion of modern agricultural technologies. Meanwhile, the “expansion coupling” type was mainly concentrated during 2004–2007, reflecting synchronous growth in carbon emissions and output value during the early expansion phase of planting activities in certain regions.

From a spatial perspective, the eastern region was dominated by “weak decoupling” and “strong decoupling”. Particularly, in economically developed provinces such as Jiangsu and Zhejiang, the “strong decoupling” type was significant, showcasing the effectiveness of advanced agricultural technologies and policy interventions. The central region was primarily characterized by “weak decoupling”, but instances of “strong decoupling” have gradually increased in recent years, indicating improved coordination between emission reductions and output value growth. In the western region, “weak decoupling” was the predominant type overall, but “expansion coupling” was observed in regions such as Tibet, reflecting the phased characteristics of regional agricultural development.

Overall, the decoupling relationship between carbon emissions and output value in the planting industry across China exhibited significant temporal evolution and spatial variation. The progression from “weak decoupling” to “strong decoupling” reflects the success of agricultural modernization and energy-saving policies, while the emergence of rare decoupling types highlights unique agricultural development patterns in specific years and regions. These findings provide valuable insights for further research on the relationship between carbon emissions and output value (Figure 4).

4.4. Analysis of Factors Influencing Carbon Emissions in the Planting Industry

The results of the dynamic panel model analysis reveal the complex driving mechanisms of carbon emissions in the planting industry. The coefficient of the lagged carbon emissions value is as high as 0.992, with a significance level of 0.01, indicating strong path dependency in carbon emissions. This path dependency is driven by multiple factors. First, agricultural production models exhibit strong inertia, as farmers tend to adhere to traditional high-input farming methods, making the transition to low-carbon practices relatively slow. Second, agricultural infrastructure investments, such as irrigation systems and mechanized equipment, have long lifespans, limiting the speed at which new, more sustainable technologies can be adopted. Third, policy implementation varies across regions, with differences in financial support, technical training, and enforcement, further contributing to regional disparities in emission trajectories. This suggests that historical levels of carbon emissions have a significant impact on current emissions. This inertia may stem from the long-term stability of agricultural production models and the persistence of resource utilization patterns. This finding highlights the need for stronger policy interventions to break the path dependency and achieve meaningful emission reductions.

Energy consumption exerts a significant promoting effect on carbon emissions, with a coefficient of 0.022 at a significance level of 0.01. A 1% increase in energy consumption approximately leads to a 0.022% increase in carbon emissions. This underscores the high dependency of agricultural production on energy, particularly in energy-intensive activities such as irrigation and mechanized farming. Optimizing energy efficiency and adopting clean energy technologies are therefore central to achieving low-carbon agriculture.

In contrast, the gross output value of agriculture, forestry, animal husbandry, and fishery has a significant inhibitory effect on carbon emissions, with a coefficient of −0.036 at a significance level of 0.01. A 1% increase in output value reduces carbon emissions by approximately 0.036%. This indicates that efficient economic growth in agriculture, driven by productivity improvements, can contribute to resource conservation and emission reductions, reflecting the combined effects of technological progress, structural optimization, and sustainable agricultural practices.

The rural population has a positive effect on carbon emissions, with a coefficient of 0.023 at a significance level of 0.01. A 1% increase in rural population leads to a 0.023% increase in carbon emissions. This may be due to the expansion of agricultural activities and the growing demand for resources such as land and energy. It suggests that rural revitalization strategies should consider the potential pressure of population growth on resources and the environment. Agricultural mechanization shows a significant negative effect on carbon emissions, with a coefficient of −0.018 at a significance level of 0.05. This indicates that increased mechanization can effectively reduce carbon emissions, likely due to improved production efficiency and optimized resource utilization. However, this overall effect may obscure heterogeneity among different types of mechanization. While modern precision agriculture equipment and electric-powered machinery are designed to improve energy efficiency, conventional high-energy-consuming machinery, such as diesel-powered tractors and large-scale harvesters, can lead to increased carbon emissions due to their reliance on fossil fuels. The short-term carbon impact of mechanization may vary depending on the energy source used. In regions where mechanization is primarily fueled by diesel, emission reductions may be limited or even offset by the additional energy consumption required to operate large-scale machinery. Conversely, in areas where renewable energy sources are integrated into agricultural mechanization, such as electric-powered tractors and AI-driven precision farming technologies, mechanization can contribute to substantial emission reductions. These findings suggest that while mechanization plays a key role in reducing carbon emissions, its benefits are highly dependent on the energy efficiency of the machinery used. Policymakers should consider not only increasing the mechanization rate but also ensuring that the transition to energy-efficient agricultural machinery is supported by appropriate incentives and infrastructure.

Cultivated land area has a significant positive effect on carbon emissions, with a coefficient of 0.021 at a significance level of 0.01. This reflects the significant role of agricultural land expansion in driving carbon emissions, particularly in the context of land reclamation and increased use of marginal lands. Rational land use planning is therefore critical for reducing emissions. Rural electricity consumption also has a significant negative effect, with a coefficient of −0.011 at a significance level of 0.05. This suggests that the adoption of electricity-driven technologies can reduce carbon intensity, as agricultural electrification replaces traditional energy sources such as fossil fuels. The integration of electrification with mechanization could further enhance emission reductions, particularly if clean energy sources are used to power agricultural equipment.

Agricultural investment has a significant positive effect on carbon emissions, with a coefficient of 0.01 at a significance level of 0.01. This indicates that increased investment in agriculture may lead to higher carbon emissions, reflecting the resource-intensive nature of inputs such as fertilizers, pesticides, and machinery. However, this finding also suggests the need to differentiate between conventional investment and green investment. Investments in clean energy, precision farming, and low-carbon mechanization technologies could mitigate these emission pressures while sustaining agricultural productivity.

In summary, the results of the dynamic panel model reveal multiple driving factors and their complex relationships with carbon emissions in the planting industry. On one hand, variables such as energy consumption, cultivated land area, and agricultural investment significantly promote carbon emissions, highlighting the close association between agricultural expansion and resource consumption. On the other hand, variables such as AFAF output, mechanization level, and rural electricity consumption significantly inhibit carbon emissions, emphasizing the critical role of technological progress and resource efficiency in emission reductions. Notably, the impact of mechanization on emissions is not uniform, as high-energy-consuming machinery can partially offset the carbon reduction benefits of mechanization. Therefore, future policies should focus on not only expanding mechanization but also transitioning toward energy-efficient and low-carbon agricultural equipment. The future focus for low-carbon agricultural development should include optimizing energy use, advancing mechanization, promoting electrification, and controlling land expansion while leveraging green investments to achieve a dual win of production efficiency and emission reduction benefits (

Table 5. Factors Influencing Carbon Emissions in the Planting Industry.

Variable	Coef	Robust Std. Err	z	p	95% CI
Constant	0.046	0.026	1.731	0.083	−0.006~0.098
L1. ln Carbon Emissions	0.992	0.018	56.462	0.000 **	0.957~1.026
ln Energy	0.022	0.005	4.299	0.000 **	0.012~0.031
ln AFAF Output	−0.036	0.005	−7.858	0.000 **	−0.045~−0.027
ln Rural Population	0.023	0.007	3.502	0.000 **	0.010~0.036
ln Mechanization Level	−0.018	0.007	−2.439	0.015 *	−0.032~−0.004
ln Cultivated Land Area	0.021	0.008	2.738	0.006 **	0.006~0.036
ln Rural Electricity Use	−0.011	0.004	−2.559	0.011 *	−0.019~−0.002
ln Agricultural Investment	0.01	0.002	4.384	0.000 **	0.006~0.014

* p < 0.05, ** p < 0.01.

).

4.5. Spatial Spillover Effects of Agricultural Carbon Emissions

Table 6. Results of the Spatial Durbin Model (SDM).

Variable	Direct Effect	Spillover Effect	Total Effect
Mechanization	−0.035 *	0.012	−0.023 *
Agricultural Investment	0.045 **	0.027 **	0.072 **
Rural Electricity	−0.019	−0.005	−0.024
Land Area	0.022 *	0.008	0.030 *
Spatial Lag	0.218 *

* p < 0.05, ** p < 0.01.

presents the estimation results of the Spatial Durbin Model (SDM). The spatial autoregressive coefficient (ρ) is significantly positive (p < 0.05), indicating strong spatial dependence in carbon emissions across regions. This suggests that carbon reduction policies in one province may indirectly influence adjacent regions through industrial linkages and technology diffusion.

Mechanization exhibits a significant negative direct effect on emissions (β_k < 0, p < 0.05), confirming its role in reducing local emissions. However, its spillover effect (θ_k > 0, p < 0.1) suggests that increased mechanization in one province may slightly raise emissions in neighboring regions, possibly due to cross-regional expansion of energy-intensive agricultural practices. This highlights the importance of coordinated regional policies to mitigate negative externalities.

Similarly, agricultural investment has a strong positive spillover effect (θ_k > 0, p < 0.01), suggesting that capital-intensive agricultural development in one region tends to drive carbon-intensive activities in adjacent areas. This underscores the need for balanced regional investment strategies to minimize unintended environmental consequences.

4.6. LMDI Model

Based on the LMDI decomposition results from 2004 to 2022, the driving factors of carbon emissions in the planting industry can be divided into two categories: positive drivers and negative inhibitors. The cumulative contributions of these factors reveal their long-term impact on carbon emissions. Overall, planting industry output, agricultural investment, and rural electricity consumption are the primary drivers of carbon emissions, while energy intensity, mechanization level, and rural population play significant roles in mitigating carbon emission growth.

The cumulative contribution of planting industry output reached 187,542.13 (10,000 tons), making it the most significant driving factor for carbon emission growth. This indicates that as the scale of the planting industry expanded and production activities intensified, carbon emissions increased substantially. Agricultural investment was the second-largest driver, with a cumulative contribution of 102,190.08 (10,000 tons), reflecting the resource-intensive nature of capital inputs such as fertilizers, pesticides, irrigation infrastructure, and machinery, which led to increased resource consumption and carbon emissions. Additionally, rural electricity consumption contributed 23,048.31 (10,000 tons). Although electrification technologies have improved energy efficiency to some extent, their widespread application also raised energy demand in agricultural production, further driving carbon emissions. Similarly, land use contributed 9980.84 (10,000 tons), highlighting the long-term impact of farmland expansion and intensive land use on carbon emissions.

In contrast, the negative inhibitors significantly curtailed carbon emissions. Energy intensity had the largest inhibitory effect, with a cumulative contribution of −130,866.28 (10,000 tons). This indicates that as agricultural technologies advanced and energy efficiency improved, carbon emissions per unit output decreased substantially, particularly after 2008, when the reduction effect became more pronounced. The cumulative contribution of the rural population was −84,265.42 (10,000 tons), reflecting the mitigating effect of the rural population decline on carbon emissions, which may be associated with urbanization and the transfer of agricultural labor. The mechanization level had a cumulative contribution of −47,517.43 (10,000 tons), indicating that the adoption of mechanization technologies improved production efficiency and optimized resource use, achieving significant emission reductions. The cumulative contribution of the energy structure was −2923.80 (10,000 tons). Although relatively small, the increase in the proportion of clean energy provided critical support for low-carbon agriculture. The cumulative contribution of AFAF output was −6343.53 (10,000 tons), reflecting the slight inhibitory effect of structural adjustments within the agricultural sector, likely due to differences in production efficiency and energy use across various agricultural sub-sectors.

From a temporal perspective, the contribution of planting industry output to carbon emissions showed an upward trend, particularly after 2010, when its driving effect became increasingly prominent. This suggests that the expansion of agricultural scale is a major source of carbon emission growth. The contribution of agricultural investment exhibited a fluctuating upward trend, indicating that the resource-intensive model consistently promoted carbon emissions, although its growth stability was influenced by external economic conditions and policy changes. On the other hand, the inhibitory effect of energy intensity on carbon emissions gradually strengthened, especially after 2008, when improvements in energy efficiency significantly advanced the agricultural emission reduction process. Furthermore, the mechanization level’s inhibitory effect became evident starting in 2004, demonstrating that the transformation of agricultural production models provided essential momentum for carbon emission reduction. The mitigating effect of rural population decline was closely linked to urbanization, which facilitated the modernization and intensification of agricultural production.

In summary, the LMDI decomposition results highlight the complex driving mechanisms of carbon emissions in the planting industry. Planting industry output and agricultural investment are the primary drivers of carbon emissions, while the energy intensity, mechanization level, and rural population are key inhibitors. These findings suggest that the future development of low-carbon agriculture should focus on balancing energy optimization, mechanization efficiency improvement, and controlled agricultural expansion. Moreover, strengthening green agricultural investment and promoting clean energy technologies will be critical to achieving agricultural emission reduction goals (

Table 7. Contributions of driving factors to carbon emissions in the planting industry factor.

Year	$∆{\mathit{C}}_{\mathit{E}}$	$∆{\mathit{C}}_{\mathit{H}}$	$∆{\mathit{C}}_{\mathit{A}}$	$∆{\mathit{C}}_{\mathit{P}}$	$∆{\mathit{C}}_{\mathit{K}}$	$∆{\mathit{C}}_{\mathit{M}}$	$∆{\mathit{C}}_{\mathit{L}}$	$∆{\mathit{C}}_{\mathit{U}}$	$∆{\mathit{C}}_{\mathit{I}}$	Total
2004	−728.13	−439.84	4.71	1553.75	−427.82	−622.13	217.81	815.04	59.77	433.15
2005	−852.73	−774.95	59.78	2163.87	−921.00	−963.86	−319.53	2137.19	162.92	691.69
2006	−708.54	−980.24	−427.98	2977.04	−1431.03	−493.34	−1129.27	3154.18	−14.42	946.40
2007	−830.95	−2062.73	−27.39	4083.64	−1894.04	−1988.00	−679.89	4641.60	32.92	1275.16
2008	−117.22	−4070.80	325.06	4699.32	−2033.61	−1712.90	−3527.77	7691.18	174.23	1427.49
2009	−196.47	−4087.32	−81.03	5566.98	−2711.58	−1769.81	−4986.99	9691.37	269.44	1694.59
2010	−223.93	−5038.99	−495.47	7477.16	−3511.36	−1943.15	−5469.55	10,831.64	347.77	1974.12
2011	−143.72	−6304.25	−190.90	8868.96	−4276.92	−2230.12	1611.79	4447.35	451.83	2234.02
2012	−149.11	−6984.20	−327.93	10,153.14	−4984.37	−2277.68	846.92	5645.67	552.98	2475.42
2013	−212.01	−7530.37	−433.12	11,194.20	−5308.58	−3355.81	2651.09	4981.33	648.32	2635.07
2014	−161.03	−7947.55	−519.14	11,986.65	−5882.90	−3376.60	3373.10	4591.24	736.07	2799.82
2015	−189.78	−8260.27	−554.30	12,633.91	−6407.73	−3227.73	3621.42	4425.97	829.70	2871.19
2016	93.47	−8976.82	−391.95	13,047.75	−5330.11	−4699.09	3374.76	4851.29	832.90	2802.19
2017	26.38	−9227.97	−531.29	13,518.93	−5632.38	−4789.79	3661.47	4797.84	792.15	2615.34
2018	136.76	−9812.10	−688.86	13,981.54	−5863.10	−4410.52	2683.78	5500.01	755.82	2283.32
2019	144.78	−10,715.00	−544.32	14,535.77	−6131.64	−4242.88	2767.00	5387.21	744.05	1944.98
2020	556.33	−12,114.60	−337.57	15,682.15	−6901.13	−4166.32	1770.84	6417.73	816.07	1723.52
2021	432.39	−12,584.80	−533.10	16,519.68	−7219.10	−816.19	−1637.66	6557.98	871.95	1591.19
2022	199.71	−12,953.50	−648.74	16,897.70	−7397.02	−431.51	14,219.00	5624.26	916.37	16,426.22
Total	−2923.80	−130,866.00	−6343.54	187,542.10	−84,265.50	−47,517.40	23,048.33	102,190.10	9980.83

).

4.7. Coupling Coordination Relationship Between Positive Driving Factors and Carbon Emissions

From 2003 to 2022, the coupling coordination relationship between positive driving factors (planting industry output, rural electricity consumption, agricultural investment, and land use) and carbon emissions evolved significantly, shifting from “imbalance” to “coordination”. Overall, as agricultural production efficiency improved and resource utilization was optimized, the positive effects of these factors on carbon emissions gradually weakened, leading to better coupling coordination. The coupling coordination relationship between planting industry output and carbon emissions was in a state of extreme imbalance (level 1) in 2003 but improved year by year, reaching barely coordinated (level 6) in 2009 and stabilizing with good coordination (level 9) from 2013 to 2022. This indicates that advancements in agricultural technology and production efficiency significantly enhanced the coordination between planting industry output and carbon emissions, reflecting a trend toward low-carbon modern agricultural production. The coupling relationship between rural electricity consumption and carbon emissions also improved significantly. It transitioned from extreme imbalance (level 1) in 2003 to primary coordination (level 7) in 2007 and achieved intermediate coordination (level 8) over the subsequent decade. This demonstrates the positive role of widespread electrification technologies in improving energy efficiency in agriculture. However, in 2021 and 2022, the coupling level slightly dropped to level 7, possibly due to the rapid growth in rural electricity demand outpacing improvements in energy structure optimization.

Agricultural investment exhibited a similar trend, starting in extreme imbalance (level 1) but rapidly improving from 2008 onward, reaching optimal coordination (level 10) by 2010. This result highlights the significant impact of optimized agricultural investment structures, particularly the promotion of green investments and low-carbon technologies, which substantially reduced the negative effects of capital inputs on carbon emissions and greatly enhanced the coordination between investment and carbon emissions. The coupling coordination relationship between land use and carbon emissions started in extreme imbalance (level 1) in 2003 but improved to barely coordinated (level 6) by 2008 and further to intermediate coordination (level 8) after 2010. This suggests that enhanced land resource utilization efficiency and measures to control farmland expansion have gradually reduced the driving force of land use on carbon emissions, leading to significant improvements in coordination.

In summary, the coupling coordination relationship between all positive driving factors and carbon emissions improved from extreme imbalance (level 1) in 2003 to intermediate coordination (level 8) in 2008 and reached good coordination (level 9) during the 2014–2020 period. This indicates that advancements in agricultural technology, improved management efficiency, and the deep integration of green development concepts have stabilized and enhanced the coordination between driving factors and carbon emissions. However, the recent decline in the coupling level of rural electricity consumption underscores the need for further optimization of the energy structure and increased adoption of clean energy while accelerating agricultural electrification. Moreover, continued efforts in promoting green agricultural investment and rational land use planning will contribute to achieving higher levels of coupling coordination, providing strong support for low-carbon agricultural development (

Table 8. Coupling coordination relationship between positive driving factors and carbon emissions.

Year	Planting Industry Output	Rural Electricity Consumption	Agricultural Investment	Land Use	All Positive Driving Factors
2003	1	1	1	1	1
2004	3	4	3	4	3
2005	4	5	5	5	4
2006	5	5	5	1	1
2007	5	7	6	4	5
2008	6	7	8	6	6
2009	6	8	9	7	7
2010	7	8	10	8	8
2011	8	9	7	8	7
2012	8	9	8	9	8
2013	9	10	8	9	8
2014	9	10	8	10	8
2015	9	10	8	10	9
2016	9	10	8	10	9
2017	9	10	8	10	9
2018	9	10	8	9	9
2019	9	9	8	9	9
2020	9	9	8	9	9
2021	9	8	8	9	9
2022	9	7	7	9	8

).

5. Discussion

This study systematically analyzed the temporal trends, carbon emission intensity, decoupling relationships, driving factor decomposition, and coupling coordination of carbon emissions in China’s planting industry, uncovering intrinsic patterns and key influencing mechanisms. These findings highlight the complex and dynamic interplay between agricultural development and carbon emissions, offering valuable policy insights for achieving a low-carbon transition in the planting industry. However, compared with other major agricultural economies, China’s progress in reducing carbon emissions from planting remains relatively slow and uneven, underscoring the need for technological and structural advancements to enhance emission reduction efficiency.

This study’s methodological approach was chosen based on the complexity and dynamic evolution of the influencing factors of agricultural carbon emissions. Compared to conventional methods, we employed models that provide stronger explanatory power and applicability.

In analyzing the driving factors of carbon emissions, traditional studies typically rely on static regression models to assess the impact of agricultural inputs, energy consumption, and other factors. However, static models fail to capture the path dependency of carbon emissions and cannot effectively identify the lagged effects of past emissions on current levels. Our study adopts a dynamic panel model (DPM), which incorporates lagged carbon emissions as an explanatory variable, providing a more accurate representation of the evolutionary mechanism of agricultural carbon emissions. Additionally, the dynamic panel model effectively mitigates endogeneity issues, thereby improving the reliability of the estimation results. In contrast, traditional static models overlook the cumulative effects over time, potentially underestimating the impact of policy interventions.

For decomposing the drivers of carbon emissions, conventional methods primarily rely on index decomposition analysis (IDA), which may generate residuals in the decomposition process, leading to unexplained changes in carbon emissions. In contrast, we applied the Logarithmic Mean Divisia Index (LMDI) decomposition method, which offers zero residual and additivity properties, ensuring that changes in carbon emissions are fully attributed to specific driving factors, thereby enhancing the accuracy of results. Additionally, LMDI is well suited for multi-dimensional and multi-variable analysis, making it a superior tool for capturing the impact of agricultural activities on carbon emissions. Conventional index decomposition approaches may have limitations when handling complex multi-factor scenarios.

In evaluating the relationship between agricultural economic growth and carbon emissions, conventional studies often use correlation analysis or single regression models. While these methods reveal the basic association between agricultural development and emissions, they fail to assess the coupling coordination between the two. Our study employed the coupling coordination model (CCM), which quantitatively evaluates the synergy between agricultural economic growth and carbon emissions while identifying regional development trends. This provides more targeted policy insights for low-carbon agricultural development. Unlike traditional regression models that only explore linear or nonlinear relationships, CCM measures the extent to which agricultural economic growth and carbon emissions evolve in a coordinated manner.

The dynamic changes in carbon emissions reveal distinct phases in China’s agricultural development. Between 2003 and 2022, carbon emissions in the planting industry initially increased, reaching a peak due to the expansion of mechanized farming, intensified use of agrochemicals, and land use changes, before subsequently declining as agricultural efficiency improvements and optimized input use contributed to emission reductions. However, the overall reduction rate remains modest, suggesting that further advancements in agricultural efficiency and sustainability practices will be necessary to align with long-term low-carbon development goals.

6. Policy Implications and Recommendations

Achieving a low-carbon transition in China’s planting industry requires a multi-faceted approach that balances economic feasibility, technological innovation, and institutional support. While the findings of this study highlight progress in decoupling carbon emissions from agricultural growth, persistent challenges such as financial barriers, technological gaps, and regional disparities must be addressed through targeted policies. To enhance the effectiveness of emission reduction efforts, this study proposes the following policy recommendations.

6.1. Enhancing Technology Adoption Through Targeted Subsidies and Financial Support

A major obstacle to low-carbon agricultural transformation is the high initial investment required for precision agriculture, renewable-powered irrigation, and low-carbon mechanization. The government should expand targeted subsidies for energy-efficient farm equipment and climate-smart agricultural technologies. For instance, an incremental subsidy model could be adopted, where farmers receive higher subsidy rates (up to 50%) for purchasing electric tractors, solar-powered irrigation systems, and AI-driven farm management tools in regions with high carbon emissions. This approach has been successfully implemented in China’s Agricultural Machinery Purchase Subsidy Program, which currently provides 15–35% subsidies for general mechanization but lacks specific incentives for low-carbon technologies.

Additionally, green financing mechanisms such as low-interest loans and agricultural carbon credit trading schemes should be expanded to incentivize private-sector investment in sustainable farming. Pilot programs in Zhejiang and Jiangsu provinces have demonstrated that offering zero-interest loans for precision irrigation systems increases adoption rates by more than 30%. Scaling up such initiatives nationwide could significantly accelerate low-carbon agricultural transformation.

6.2. Balancing Carbon Reduction with Economic and Social Trade-Offs

Transitioning to low-carbon agriculture involves inherent trade-offs between emission reduction and economic viability. Policies should ensure that carbon mitigation efforts do not disproportionately impact smallholder farmers, who may lack the financial capacity to transition away from high-input farming practices. One solution is to establish regional agricultural innovation hubs, where small-scale farmers can access shared resources such as AI-powered monitoring systems, no-till seeders, and biogas digesters. These hubs could be co-financed by government grants, private agribusinesses, and state-owned enterprises, ensuring equitable access to modern farming techniques.

Additionally, concerns over food security must be addressed. Reducing fertilizer and pesticide use can lower emissions but may also lead to short-term yield reductions. To mitigate this risk, policies should focus on integrated soil fertility management, combining organic fertilizers, microbial inoculants, and precision nutrient application to maintain crop yields while cutting emissions. Evidence from Japan’s sustainable rice farming programs suggests that such strategies can reduce emissions by 20–30% without affecting productivity.

6.3. The Role of State-Owned Agribusinesses in Low-Carbon Agriculture

State-owned agribusinesses such as COFCO Group and Beidahuang Agricultural Reclamation Group play a crucial role in China’s agricultural supply chain. These enterprises should be integrated into the national carbon reduction strategy by adopting large-scale demonstration farms that showcase best practices in low-carbon cultivation. For example, COFCO’s green supply chain initiative, which promotes sustainable sourcing and carbon footprint tracking, could be expanded to incentivize contract farmers to adopt climate-smart practices in exchange for premium pricing or preferential procurement agreements.

Additionally, agricultural carbon markets should include large agribusinesses as key participants. Pilot programs in China’s forestry sector have demonstrated the feasibility of carbon credit trading, where landowners receive financial compensation for maintaining carbon-sequestering ecosystems. A similar system could be applied to agriculture, where large-scale farms and cooperatives earn carbon credits for adopting conservation tillage, methane capture, or solar-powered irrigation.

6.4. Strengthening Regional Policy Coordination for Equitable Decarbonization

Decarbonization efforts must account for regional economic and technological disparities. While the eastern provinces have successfully implemented precision agriculture and digitalized farm management, the central and western regions still rely on traditional farming practices with lower resource efficiency. A “National Low-Carbon Agriculture Collaboration Mechanism” should be established to facilitate technology transfer and knowledge sharing between developed and underdeveloped regions. This could include cross-regional training programs, government incentives for agribusinesses that establish joint ventures in western China, and preferential tax policies for agricultural companies investing in sustainable farming in lower-income areas.

Expanding China’s rural revitalization strategy to include region-specific low-carbon policies will also be essential. Targeted interventions, such as increasing subsidy rates for water-saving irrigation technologies in arid regions and prioritizing carbon credit programs in high-emission farming zones, can help create a balanced and equitable transition to sustainable agriculture.

7. Conclusions

This study reveals the complex dynamics of carbon emissions in China’s planting industry, highlighting both the progress made and the challenges that persist in achieving sustainable agricultural development. While technological advancements and policy interventions have contributed to improvements in emission efficiency, the overall reduction trajectory remains slow and uneven. The findings indicate that regional disparities in agricultural modernization and resource utilization continue to hinder nationwide carbon mitigation efforts, necessitating a more coordinated and region-specific policy approach. The expansion of the planting industry and agricultural investment remain the primary drivers of carbon emissions, yet the disproportionate allocation of resources toward high-input farming methods exacerbates environmental pressures. Conversely, improvements in energy efficiency and mechanization demonstrate the potential to significantly reduce emissions, particularly in regions where clean energy and digital agriculture have been effectively integrated into production systems.

Looking forward, China must accelerate the transition to low-carbon agriculture by prioritizing green technology adoption, optimizing resource allocation, and promoting more balanced regional development policies. Strengthening policy coordination between national and local governments will be critical to ensuring that emission reduction strategies are effectively implemented across diverse agricultural landscapes. Financial and technological support should be expanded, particularly for central and western regions, to narrow the technological gap and facilitate the adoption of modern, low-emission farming practices. The significance of this study lies in its comprehensive analysis of the current state and driving mechanisms of carbon emissions in the planting industry, offering not only theoretical contributions but also practical policy recommendations for enhancing carbon reduction strategies.

Achieving low-carbon development in the planting industry requires not only technological innovation and policy intervention but also structural transformations in regional coordination, land use management, and energy transition strategies. This effort is not only crucial for the sustainable development of China’s agriculture but also provides valuable insights for other rapidly developing economies seeking to balance agricultural productivity with environmental sustainability. Strengthening international cooperation in sustainable agriculture, knowledge-sharing on emission reduction technologies, and collaborative policy frameworks will further enhance China’s ability to integrate agricultural modernization with carbon neutrality goals.